双臂对称性对压缩传感用于关联成像重构的影响  

Influence of two-arm symmetry on reconstructed image of compressive sensing for ghost imaging

作　　者：王铭海[1] 曹军胜[2] 郜峰利[1] 

机构地区：[1]吉林大学电子科学与工程学院集成光电子学国家重点联合实验室吉林大学实验区,吉林长春130012 [2]中国科学院长春光学精密机械与物理研究所,吉林长春130033

出　　处：《光学精密工程》2014年第6期1438-1445,共8页Optics and Precision Engineering

基　　金：国家自然科学基金青年基金资助项目(No.61204055);吉林省科技发展计划青年科研基金资助项目(No.20130522188JH)

摘　　要：根据压缩传感理论和关联成像模型,将压缩传感理论应用于关联成像中,实现了传统的双臂关联成像的压缩传感重构。通过仿真实验验证了压缩传感用于关联成像的可行性,以峰值信噪比(PSNR)为衡量指标,分别对压缩传感和传统关联算法的重构图像质量进行了量化。仿真实验表明,压缩传感和关联算法的重构效果均随测量次数的增加而变优,在相同的测量次数下,压缩传感在关联成像中的重构图像的PSNR比传统的关联重构图像高20dB以上。将压缩传感用于实际双臂关联成像的实验结果表明,压缩传感可以实现双臂关联成像装置的图像重构,但其重构质量很难优于传统关联算法的重构。针对这一实际实验与仿真实验似乎相矛盾的特殊现象,从双臂对称性的角度进行了合理解释,并利用实验中实际的散斑场对该现象进行了验证,最后提出了解决方案。According to Compressive Sensing （CS） algorithms and the Ghost imaging （GI） model, the CS was applied to the GI system to complete the CS reconstruction of an image. The feasibility of CS applied in GI was firstly validated by a simulation experiment. By using Peak Signal to Noise Ratio （PSNR） as the measure, the restructured images based on CS and traditional GI correlation algorithms were quantified respectively. The simulation experiment results indicate that both the restructured images are getting better with the increase of the number of measurements, however, the PSNR of CS reconstruction image is above 20 dB higher than that of the traditional correlation reconstruction method at the same number of measurements. Furthermore, the CS was applied in an actual two-armGI experiment. The experiment results indicate that the CS can achieve the image reconstruction of two-arm correlation imaging equipment, but its reconstruction quality is hard to be better than that of the GI correlation algorithm. For this special confliction phenomenon, the paper gives some reasona- ble interpretations from the two-arm symmetry perspective and then fully validates the interpretations by using the actual speckle pattern from the experiment. Finally, it proposes a solution scheme.

关 键 词：关联成像 双臂关联成像装置 双臂对称 压缩传感 稀疏表示 图像重构 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN27[自动化与计算机技术—计算机科学与技术]